re PR tree-optimization/69848 (poor vectorization of a loop from SPEC2006 464.h264ref)
authorBin Cheng <bin.cheng@arm.com>
Tue, 16 Aug 2016 13:09:40 +0000 (13:09 +0000)
committerBin Cheng <amker@gcc.gnu.org>
Tue, 16 Aug 2016 13:09:40 +0000 (13:09 +0000)
commit3c556bc4e9eb31561f246e8e2944b05a95cc1a4a
tree3f3f8f1a9f4e26388ee99a4cec27390b5ef38284
parentedd1a1cb736d0b32396633f2ff45de46a33efb74
re PR tree-optimization/69848 (poor vectorization of a loop from SPEC2006 464.h264ref)

PR tree-optimization/69848
* config/aarch64/aarch64-simd.md (vcond<mode><mode>): Invert NE
and swtich operands to avoid additional NOT instruction.
(vcond<v_cmp_mixed><mode>): Ditto.
(vcondu<mode><mode>, vcondu<mode><v_cmp_mixed>): Ditto.

gcc/testsuite
* gcc.target/aarch64/simd/vcond-ne-bit.c: New test.

From-SVN: r239502
gcc/ChangeLog
gcc/config/aarch64/aarch64-simd.md
gcc/testsuite/ChangeLog
gcc/testsuite/gcc.target/aarch64/simd/vcond-ne-bit.c [new file with mode: 0644]